Commit @3e0223249375bf1a38bb80eff4370510881c1b93 - yjyoon/whisper_streaming

Alex Young 2024-04-18

Merge branch 'main' into ayo-logging-fixes

@3e0223249375bf1a38bb80eff4370510881c1b93

259a346

3e02232

README.md

--- README.md

+++ README.md


 
 ### Server -- real-time from mic
 
-`whisper_online_server.py` has the same model options as `whisper_online.py`, plus `--host` and `--port` of the TCP connection. See help message (`-h` option).
+`whisper_online_server.py` has the same model options as `whisper_online.py`, plus `--host` and `--port` of the TCP connection and the `--warmup-file`. See the help message (`-h` option).
 
 Client example:
 

259a346

3e02232

line_packet.py

--- line_packet.py

+++ line_packet.py


 
 """Functions for sending and receiving individual lines of text over a socket.
 
-Used by marian-server-server.py to communicate with the Marian worker.
-
 A line is transmitted using one or more fixed-size packets of UTF-8 bytes
 containing:
 

 
   - Zero or more \0 bytes as required to pad the packet to PACKET_SIZE
 
+Originally from the UEDIN team of the ELITR project. 
 """
 
 PACKET_SIZE = 65536

259a346

3e02232

whisper_online.py

--- whisper_online.py

+++ whisper_online.py


 
 def asr_factory(args, logfile=sys.stderr):
     """
-    Creates and configures an ASR instance based on the specified backend and arguments.
+    Creates and configures an ASR and ASR Online instance based on the specified backend and arguments.
     """
     backend = args.backend
     if backend == "openai-api":

         logging.info("Setting VAD filter")
         asr.use_vad()
 
-    return asr
+    language = args.lan
+    if args.task == "translate":
+        asr.set_translate_task()
+        tgt_language = "en"  # Whisper translates into English
+    else:
+        tgt_language = language  # Whisper transcribes in this language
 
+    # Create the tokenizer
+    if args.buffer_trimming == "sentence":
+        tokenizer = create_tokenizer(tgt_language)
+    else:
+        tokenizer = None
+
+    # Create the OnlineASRProcessor
+    online = OnlineASRProcessor(asr,tokenizer,logfile=logfile,buffer_trimming=(args.buffer_trimming, args.buffer_trimming_sec))
+
+    return asr, online
 ## main:
 
 if __name__ == "__main__":

     duration = len(load_audio(audio_path))/SAMPLING_RATE
     logging.info("Audio duration is: %2.2f seconds" % duration)
 
-    asr = asr_factory(args, logfile=logfile)
-    language = args.lan
-
-    if args.task == "translate":
-        asr.set_translate_task()
-        tgt_language = "en"  # Whisper translates into English
-    else:
-        tgt_language = language  # Whisper transcribes in this language
-
+    asr, online = asr_factory(args, logfile=logfile)
     min_chunk = args.min_chunk_size
-    if args.buffer_trimming == "sentence":
-        tokenizer = create_tokenizer(tgt_language)
-    else:
-        tokenizer = None
-    online = OnlineASRProcessor(asr,tokenizer,logfile=logfile,buffer_trimming=(args.buffer_trimming, args.buffer_trimming_sec))
-
 
     # load the audio into the LRU cache before we start the timer
     a = load_audio_chunk(audio_path,0,1)
 
-    # warm up the ASR, because the very first transcribe takes much more time than the other
+    # warm up the ASR because the very first transcribe takes much more time than the other
     asr.transcribe(a)
 
     beg = args.start_at

259a346

3e02232

whisper_online_server.py

--- whisper_online_server.py

+++ whisper_online_server.py


 # server options
 parser.add_argument("--host", type=str, default='localhost')
 parser.add_argument("--port", type=int, default=43007)
+parser.add_argument("--warmup-file", type=str, dest="warmup_file", 
+        help="The path to a speech audio wav file to warm up Whisper so that the very first chunk processing is fast. It can be e.g. https://github.com/ggerganov/whisper.cpp/raw/master/samples/jfk.wav .")
 
 parser.add_argument("-l", "--log-level", dest="log_level", 
                     choices=['DEBUG', 'INFO', 'WARNING', 'ERROR', 'CRITICAL'],

 
 size = args.model
 language = args.lan
-
-asr = asr_factory(args)
-
-if args.task == "translate":
-    asr.set_translate_task()
-    tgt_language = "en"
-else:
-    tgt_language = language
-
+asr, online = asr_factory(args)
 min_chunk = args.min_chunk_size
+
 
 if args.buffer_trimming == "sentence":
     tokenizer = create_tokenizer(tgt_language)

     tokenizer = None
 online = OnlineASRProcessor(asr,tokenizer,buffer_trimming=(args.buffer_trimming, args.buffer_trimming_sec))
 
-
-
-demo_audio_path = "cs-maji-2.16k.wav"
-if os.path.exists(demo_audio_path):
-    # load the audio into the LRU cache before we start the timer
-    logging.debug(f"Warming up on {demo_audio_path}")
-    a = load_audio_chunk(demo_audio_path,0,1)
-
-    # TODO: it should be tested whether it's meaningful
-    # warm up the ASR, because the very first transcribe takes much more time than the other
-    asr.transcribe(a)
-    logging.debug("Whisper is warmed up")
+# warm up the ASR because the very first transcribe takes more time than the others. 
+# Test results in https://github.com/ufal/whisper_streaming/pull/81
+msg = "Whisper is not warmed up. The first chunk processing may take longer."
+if args.warmup_file:
+    if os.path.isfile(args.warmup_file):
+        a = load_audio_chunk(args.warmup_file,0,1)
+        asr.transcribe(a)
+        print("INFO: Whisper is warmed up.",file=sys.stderr)
+    else:
+        print("WARNING: The warm up file is not available. "+msg,file=sys.stderr)
 else:
-    logging.debug("Whisper is not warmed up")
+    print("WARNING: " + msg, file=sys.stderr)
 
 
 ######### Server objects

Add a comment

Open 0
Closed 0

List

...	...	@@ -183,7 +183,7 @@
183	183
184	184	### Server -- real-time from mic
185	185
186		-`whisper_online_server.py` has the same model options as `whisper_online.py`, plus `--host` and `--port` of the TCP connection. See help message (`-h` option).
	186	+`whisper_online_server.py` has the same model options as `whisper_online.py`, plus `--host` and `--port` of the TCP connection and the `--warmup-file`. See the help message (`-h` option).
187	187
188	188	Client example:
189	189

...	...	@@ -2,8 +2,6 @@
2	2
3	3	"""Functions for sending and receiving individual lines of text over a socket.
4	4
5		-Used by marian-server-server.py to communicate with the Marian worker.
6		-
7	5	A line is transmitted using one or more fixed-size packets of UTF-8 bytes
8	6	containing:
9	7
...	...	@@ -11,6 +9,7 @@
11	9
12	10	- Zero or more \0 bytes as required to pad the packet to PACKET_SIZE
13	11
	12	+Originally from the UEDIN team of the ELITR project.
14	13	"""
15	14
16	15	PACKET_SIZE = 65536

...	...	@@ -559,7 +559,7 @@
559	559
560	560	def asr_factory(args, logfile=sys.stderr):
561	561	"""
562		- Creates and configures an ASR instance based on the specified backend and arguments.
	562	+ Creates and configures an ASR and ASR Online instance based on the specified backend and arguments.
563	563	"""
564	564	backend = args.backend
565	565	if backend == "openai-api":
...	...	@@ -584,8 +584,23 @@
584	584	logging.info("Setting VAD filter")
585	585	asr.use_vad()
586	586
587		- return asr
	587	+ language = args.lan
	588	+ if args.task == "translate":
	589	+ asr.set_translate_task()
	590	+ tgt_language = "en" # Whisper translates into English
	591	+ else:
	592	+ tgt_language = language # Whisper transcribes in this language
588	593
	594	+ # Create the tokenizer
	595	+ if args.buffer_trimming == "sentence":
	596	+ tokenizer = create_tokenizer(tgt_language)
	597	+ else:
	598	+ tokenizer = None
	599	+
	600	+ # Create the OnlineASRProcessor
	601	+ online = OnlineASRProcessor(asr,tokenizer,logfile=logfile,buffer_trimming=(args.buffer_trimming, args.buffer_trimming_sec))
	602	+
	603	+ return asr, online
589	604	## main:
590	605
591	606	if __name__ == "__main__":
...	...	@@ -613,27 +628,13 @@
613	628	duration = len(load_audio(audio_path))/SAMPLING_RATE
614	629	logging.info("Audio duration is: %2.2f seconds" % duration)
615	630
616		- asr = asr_factory(args, logfile=logfile)
617		- language = args.lan
618		-
619		- if args.task == "translate":
620		- asr.set_translate_task()
621		- tgt_language = "en" # Whisper translates into English
622		- else:
623		- tgt_language = language # Whisper transcribes in this language
624		-
	631	+ asr, online = asr_factory(args, logfile=logfile)
625	632	min_chunk = args.min_chunk_size
626		- if args.buffer_trimming == "sentence":
627		- tokenizer = create_tokenizer(tgt_language)
628		- else:
629		- tokenizer = None
630		- online = OnlineASRProcessor(asr,tokenizer,logfile=logfile,buffer_trimming=(args.buffer_trimming, args.buffer_trimming_sec))
631		-
632	633
633	634	# load the audio into the LRU cache before we start the timer
634	635	a = load_audio_chunk(audio_path,0,1)
635	636
636		- # warm up the ASR, because the very first transcribe takes much more time than the other
	637	+ # warm up the ASR because the very first transcribe takes much more time than the other
637	638	asr.transcribe(a)
638	639
639	640	beg = args.start_at

...	...	@@ -12,6 +12,8 @@
12	12	# server options
13	13	parser.add_argument("--host", type=str, default='localhost')
14	14	parser.add_argument("--port", type=int, default=43007)
	15	+parser.add_argument("--warmup-file", type=str, dest="warmup_file",
	16	+ help="The path to a speech audio wav file to warm up Whisper so that the very first chunk processing is fast. It can be e.g. https://github.com/ggerganov/whisper.cpp/raw/master/samples/jfk.wav .")
15	17
16	18	parser.add_argument("-l", "--log-level", dest="log_level",
17	19	choices=['DEBUG', 'INFO', 'WARNING', 'ERROR', 'CRITICAL'],
...	...	@@ -33,16 +35,9 @@
33	35
34	36	size = args.model
35	37	language = args.lan
36		-
37		-asr = asr_factory(args)
38		-
39		-if args.task == "translate":
40		- asr.set_translate_task()
41		- tgt_language = "en"
42		-else:
43		- tgt_language = language
44		-
	38	+asr, online = asr_factory(args)
45	39	min_chunk = args.min_chunk_size
	40	+
46	41
47	42	if args.buffer_trimming == "sentence":
48	43	tokenizer = create_tokenizer(tgt_language)
...	...	@@ -50,20 +45,18 @@
50	45	tokenizer = None
51	46	online = OnlineASRProcessor(asr,tokenizer,buffer_trimming=(args.buffer_trimming, args.buffer_trimming_sec))
52	47
53		-
54		-
55		-demo_audio_path = "cs-maji-2.16k.wav"
56		-if os.path.exists(demo_audio_path):
57		- # load the audio into the LRU cache before we start the timer
58		- logging.debug(f"Warming up on {demo_audio_path}")
59		- a = load_audio_chunk(demo_audio_path,0,1)
60		-
61		- # TODO: it should be tested whether it's meaningful
62		- # warm up the ASR, because the very first transcribe takes much more time than the other
63		- asr.transcribe(a)
64		- logging.debug("Whisper is warmed up")
	48	+# warm up the ASR because the very first transcribe takes more time than the others.
	49	+# Test results in https://github.com/ufal/whisper_streaming/pull/81
	50	+msg = "Whisper is not warmed up. The first chunk processing may take longer."
	51	+if args.warmup_file:
	52	+ if os.path.isfile(args.warmup_file):
	53	+ a = load_audio_chunk(args.warmup_file,0,1)
	54	+ asr.transcribe(a)
	55	+ print("INFO: Whisper is warmed up.",file=sys.stderr)
	56	+ else:
	57	+ print("WARNING: The warm up file is not available. "+msg,file=sys.stderr)
65	58	else:
66		- logging.debug("Whisper is not warmed up")
	59	+ print("WARNING: " + msg, file=sys.stderr)
67	60
68	61
69	62	######### Server objects

Delete comment